October 2018

Michael Stokes (CEO, Waveform Communications; author, The Waveform Model of Vowel Perception and Production) summarized his progress since 2012 in developing a method of using waveforms and spectrograms to identify vowels in human speech.

Mr. Stokes’ model is currently the best for describing how people recognize spoken vowels.  A distinct set of vowels characterizes each language.  For example, English has ten vowels, Spanish five.  Each vowel of a language has a characteristic sound that underlies all dialects and individual pronunciations.  That characteristic sound can be visualized as spectrograms of three sound wave frequencies.  Although unpatentable as is (patent applications 14/223304 and 13/241780), a computer program (Elbow) based on the model accurately predicts spoken vowels from observations of the spectrograms.  A second program (Cobweb), especially useful to athletes, uses speech patterns to diagnose concussions in real time.

The model can be used to rapidly improve speech.  Mr. Stokes analyzes spectrograms of someone whose speech has been impaired by a stroke or by postoperative delirium or who is learning English as a second language.  Then he coaches the person on how to change the way s/he moves lips, tongue, and jaw to pronounce vowels correctly.

This model might also contribute to cross-species communication.  Humpback whales speak with vowels.  Learning how to pronounce their language(s) might help people learn how to warn them of the danger of approaching ships.

Thank you, Mr. Stokes, for sharing your interesting model with us!